Daily numbers, graphs, analysis thread 4

[Barracker]

Welcome to thread 4 of the daily updates.

Resource links:
Worldometer UK page
Financial Times Daily updates and graphs
HSJ Coronavirus updates
Johns Hopkins Coronavirus Resource Centre
NHS England stats, including breakdown by Hospital Trust
Covidly.com to filter graphs using selected data filters
ONS statistics for CV related deaths outside hospitals, released weekly each Tuesday

Thank you to all contributors for their factual, data driven, and civil discussions.

Maybe not but the more people have had it then crudely, the rate of deaths per infections is lower than it appears.

What strikes me about the death figures from France spain italy and the uk is that fundamentally the size and populations of the countrys is in the big scheme of things roughly the same and getting on for a quarter of a billion people in total.
All 4 countrys if you take out the french care home figures are remarkably simlilar statistics, haven't all 4 peaked roughly around 900 for a few days, what would be the chances of them being that similar when considering the differences/ nuances between each countrys systems/ demographics/ different lockdowns when dealing with 250000000 people, if you was to put odds on it beforehand very low I would of thought.
It seems it just explains that each countrys nhs system combined with the effects of the virus are coping with this virtually identically and the virus is dictating the figures within hospitals as they are fundamentally similar in all 4 countrys and not the governments, therefore the only true comparisons will come from knowing how many people have died outside of Hospitals.

venezia Death rates for any disease are only calculated wrt the confirmed cases, not including unknown cases - because ntbo their number is unknown

e.g. flu has a death rate of 0.02 - 0.1% (depending on the flu virus each year)
But that is only wrt the know cases of flu
The unknowns remain .... unknown

Yes I understand that. Of course it’s unknown but it’s likely that more people have had it then we know. That’s a simple point. I’m not an idiot.

Spain is more affected than Italy, France and the UK. See this chart for death rates over time normalized by population. It is taken from here: 91-divoc.com/pages/covid-visualization/

The countries have had different death rates and timing, depending on:

what measures their governments took and when,
the capability / resources of their health services,
what % of cases came within a short time period
population density
age demographics
culture e.g. multigenerational living, amount of hugging & kissing outside the family
etc

Probably the standard way to compare countries is to look at the curve for cumulative deaths,

The FT graphs - and their geek analyst - say that the USA has the steepest death curve
but also that the UK curve is steeper than Italy's

We also see that Italy and Spain have flattened out and started to drop - they are past their peak
Germany is levelling off at a much lower level
and S Korea levelled off at a much lower level than them all
(we don't belive China's curves)

Btw...I didn’t mean you were saying I was an idiot.

"Spain is more affected than Italy, France and the UK. "

That refers more to impact of the crisis and how quickly a country can recover afterwards
not to the spread of the epidemic

There is good reason that epidemiologists and virologists track the death numbers and case numbers,
without normalisation wrt population size

As the FT says, normalisation artificially makes big countries look better and small countries better

Quite easy to look up (2019 figures)

Population of Germany: 83.02 million

Population of France: 66.99 million

Population of United Kingdom: 66.65 million

Population of Italy: 60.55 million

Population of Spain: 47.07 million

@NewAccountForCorona. Sending you a hug xxx

Normalising epidemic deaths wrt population is usually something that politicians do afterwards

either leaders of larger countries trying to downplay a disaster

• or for COVID, maybe justifying why they end lockdown -

or smaller countries lobbying for more aid

I agree, BigChocFrenzy. The normalized numbers for New York and the US as a whole will look very differently. Even in small countries there are hotspots that will be much more affected than other areas.

But overall I do think that it is useful to have both absolute and normalized numbers since the absolute one will have the opposite effect and will make bigger countries look worse than smaller ones.

Is it BigChoc? what is your expertise in this field? (sorry if I have missed earlier as I was late to catch up with this thread) but find it all fascinating. And a good example of what Mumsnet does well when they do it well! I hope all the journos and policy makers are watching and taking notes .

A better explanation than I can give, for not using normalised curves,
by the FT statistics geek John Burn-Murdoch:

mobile.twitter.com/jburnmurdoch/status/1244380095164420101

People might find this interesting. www.weforum.org/agenda/2020/04/it-was-the-saltshaker-how-germany-meticulously-traced-its-coronavirus-outbreak/?fbclid=IwAR1chC3YBD6SDiqFc2ROlqqo7vtRvJl4vHZ04-nWuzIfnGIkeWVfxaCYiO0

Death rates for any disease are only calculated wrt the confirmed cases, not including unknown cases - because ntbo their number is unknown grin

No, that's not the case.

You have a 'case fatality ratio', which is the rate of fatalities to cases, where 'cases' are people who are both infected and diagnoses with infection, and which in some places due to pisspoor testing has been over 50%. (Note, again that not every country or region has the same definition of 'case', in that you might report someone as a 'case' in the following cases:

1. if they have a positive lab test
2. if they have a positive rapid test
3. if they have the clinical signs

There were people dying of corona with the clinical signs in February in many countries who were not 'cases' because it was not tested for)

Then you have 'infection fatality ratio', which is the chance of dying if you are infected with a given disease, and which is more useful to us as an individual if we are wondering 'what is the chance I will die of covid-19?'

In some cases the 'case fatality ratio' and 'infection fatality ratio' will be the same thing, in that certain very nasty diseases will always become 'cases', where as for flu, covid-19, etc., most 'infections' will not become 'cases'.

We can calculate the 'infection fatality ratio' for covid-19 (potentially) by looking for antibodies showing infections. In that case we can survey a random sample of the population and work out the % that were infected and from that come up with a number.

There are many estimates of IFR for covid-19, e.g.,

www.thelancet.com/journals/laninf/article/PIIS1473-3099(20)30243-7/fulltext

These might be wrong, but it's definitely on the agenda to find out!

I'd be so grateful if people could state their qualifications/areas of expertise or whether they are just a lay person reading links. I do know not to trust everything people say on the internet, but still it makes a difference to me.

"As the FT says, normalisation artificially makes big countries look better and small countries better"

It's in no possible way artificial to say that 10,000 deaths in, say India, is no big deal, but 10,000 deaths in Anguilla would be unspeakably horrific.

Also what JBM is talking about is not deaths but infections. Infection statistics are a bit useless to start with. When he claims that Iceland and Luxembourg go top, he is wrong.

Iceland has declared 8 covid-19 deaths while the Uk with 183x the population 1325x more deaths.

Clearly if you test the entire population of, say, Iceland, then you get more infections than Indonesia does by (checks notes) testing fewer 20,000.

Equally it matters if you want to know if Sweden's measures are effective, then you need to first and foremost before anything else normalise deaths for population.

Forget infections. Testing criteria vary dramatically. Indonesia is probably only testing politicians and their families. The UK is testing people spluttering in hospital and NHS staff. It's meaningless.

I'd be so grateful if people could state their qualifications/areas of expertise or whether they are just a lay person reading links.

Assume that everyone is a lay person unless they specifically state otherwise. And even then they might be talking bollocks.

So, in general, read what people have to say, read any offered sources, and then decide whether what they are saying makes sense. If they don't offer sources then give their opinion less weight.

Shoots The % death rate is wrt all those cases where the disease has been confirmed in some way by a professional

The definition may differ according to country, but from what you say the UK definition is:

1. if they have a positive lab test
2. if they have a positive rapid test
3. if they have the clinical signs

My point is that the death rate is not a % of a number that includes those unknown cases where people had no symptoms and no tests

pp were talking about the COVID death rate being much lower because of all the people who don't know they have it
So none of the 3 possibilities above

Unless the UK has a completely different method to Germany,
then these unknown cases are not considered in any way, when for example we calculate our % fatality rate for flu, which is what many people use to compare with the COVID fatality rate

If someone posts their real name, business email and photo,
then you could check up on them in their professional directory, look for published papers

Otherwise, assume everyone is no more qualified than you are

Comparing the number of cases is of limited use when countries have such different test programs,
but it at least gives some feel for when the curve of one country is flattening, even if the level is difficult to compare

Germany, with by far the largest population in the EEA, could obtain even apparently lower death rates if the numbers were normalised
Here, such comparisons would be regarded as a political trick, other than as secondary data

To see how we are progressing, we compare our numbers to our European neighbours, even though they have much smaller population size

Our epidemiologists have explained why the absolute numbers are used when the epidemic is still spreading

I haven't seen Imperial or other govt advisers yet use the deaths / million, so the normalisation does not seem the standard method in the Uk either

Of course normalisation wrt population matters when it comes to impact on a country

In fact, it is being discsussed here that
German total deaths may actually be lower during our period of lockdown than in the same period in 2019,
due to the decrease in death from accidents due to traffic, work, sport, alcohol - and of course seasonal flu !

That would be .... bizarre

Most countries not incuding care home deaths may lead to significant underestimation of COVID deaths

ltccovid.org/2020/04/12/mortality-associated-with-covid-19-outbreaks-in-care-homes-early-international-evidence/

"Data from 5 European countries suggest that care home residents have so far accounted for between 42% and 57% of all deaths related to COVID-19."

The definition may differ according to country, but from what you say the UK definition is:

1. if they have a positive lab test
2. if they have a positive rapid test
3. if they have the clinical signs

I'm not saying that. I'm saying that some countries will use one or more of the above. I understand for example that some countries will ONLY include a positive lab test in their figures, even if they are already dead with apparent corona (though post-mortem results will be added, if they exist). And that some countries have a system of rapid test first then lab test later

That makes it hard to compare.....

then these unknown cases are not considered in any way, when for example we calculate our % fatality rate for flu, which is what many people use to compare with the COVID fatality rate

Depends. There are PLENTY of studies estimating annual infection incidence for flu. E.g., around 10%.

www.ncbi.nlm.nih.gov/pmc/articles/PMC5934309/

AIR the elderly are less likely to get it in the UK, because of vaccination and because of reduces social contact.

Clearly if you take the annual deaths of say 80+ from flu and divide by 10% (infection rate) of the population 80+, you get a good estimate for the infection fatality rate for that, and say, the risk of going to see Granny while you have the flu.

If we take that there are 13 million under 18s, and say 13 (or something like that, AIR) deaths from flu from the same, then we can clearly see that flu is not a life-threatening disease to the under 18s.

We can quickly come up with approximate death risks from infection from flu by age by looking at the annual flu death stats reported by ONS and comparing to population and using a reasonable estimate for annual infection incidence.

Flu overwhelmingly afflicts 80+, coronavirus skews considerably younger but still doesn't bother children to any relevant degree.

Thx bimbleboo - most of the time I can be distanced and pragmatic about this and then every so often it hits me like a train and I'm simultaneously furious and gutted. But this isn't the thread, I should go and rant on the many other non-factual threads, there's loads of choice!

Chicchic, I'm just a lay person reading links, sorry, though I have a scientific background so I can usually follow links. Much of my understanding comes from two front line medical and one academic relative with whom I'm having a parallel conversation. I'm assuming everyone is a lay person with an interest in facts and figures.